Background Modeling Algorithm Based on Transitions Intensities

Abstract

Moving objects detection in a scene from a video sequence is a fundamental and crucial problem in many vision systems such as video surveillance or the traffic monitoring. Background modeling is often used to detect moving objects which assumes that the moving objects take less time in the scene with respect to the background. Indeed, this assumption cannot be valid in case of a scene with fast moving objects. To overcome this problem, in this paper we propose a novel background modeling approach based on transitions intensities. In other words, we will focus on intensity changes (transitions) of pixels during a scene instead of using duration information that a pixel pass in a state without any notable change in its intensity. The normalized RGB color is used to describe the appearance under change lighting conditions, and transition intensity is introduced to capture the dynamicity of objects over time. Experimental results on different video sequences have been presented to approve the effectiveness of the proposed algorithm.